PSO-based Spatial Data Clustering Model And Its Application

Spatial data clustering is an important part of spatial data mining. Clusteranalysis to the data objects is grouped according to the similarity. In order todiscover the law of the distribution of spatial data and its typical pattern, when inthe higher dimension of spatial data, data clustering will produce the "dimensiondisaster". According to the spatial data "dimension disaster" and the defect ofK-means clustering algorithm, on the basis of previous studies, we find out thebetter dimensionality reduction method, and improve the K-means algorithm.Then combine the PSO algorithm with the K-means algorithm, finally, improvedalgorithm is tested on multiple data sets, we get a reliable conclusion, it providesnew ideas for the improvement of spatial data clustering algorithm.In the first chapter, we elaborate the importance of spatial data clusteringalgorithm in dealing with spatial data of mass data, and explain inlow-dimensional data clustering, the practical significance of the K-meansclustering algorithm, then propose the theoretical and practical significance of theCombination of PSO and K-means clustering algorithm in the processing ofhigh-dimensional data, we describe these two clustering algorithms’ researchbackground, significance, and research status, and put forward the main work ofthis paper.In the second chapter, we have studied the K-means clustering algorithm inlow-dimensional space, first to establish a model for the K-means clusteringalgorithm, and then improve the initial centers and its evaluation function, theimproved algorithm is applied to the Iris and Wine data sets, we obtain thealgorithm which has a higher quality and stability.In the third chapter, we study spatial data PSO algorithm combined with theK-means algorithm clustering algorithm, firstly, put inertia weight in PSOclustering algorithm, secondly, PSO algorithm optimization is used to optimize theK-means algorithm, then add weights to function evaluation criterion, thirdly, it istested in Chinese environmental statistics and other data sets, we get a higheraccuracy rate algorithm.